A self-consistent method for daytime F region modeling was applied to Millstone Hill radar observations during the severe geomagnetic storm of April 6--12, 1990, when the F2 layer maximum practically disappeared at usual heights and the F1 layer constituted the ionospheric maximum during daytime hours. Neutral composition and temperature changes are shown to be the main reason for the observed tenfold decrease of electron concentration at the F2 layer heights. The decrease in [O> at 300 km is more than a factor of 6, and the [N2> and [O2> increases are 3 and 16 times, respectively, with respect to the prestorm quiet time level. Such changes of neutral composition lead to a strong decrease in O+ production rate and increase in O+ loss rate resulting in a complete disappearance of the F2 layer maximum. Our calculations for 300 km altitude indicate a nineteenfold O/N2 ratio decrease for April 10 with respect to quiet day of April 7. Horizontal plasma transfer due to the observed E-W drift is insufficient to account for the observed density changes during daytime hours. The calculated exospheric temperatures Tex are close to both the mass spectrometer incoherent scatterer (MSIS) 83 model predictions and the Millstone Hill estimates for all days except for April 10, when Tex is higher than 2000 ¿K, which may be attributed to Joule heating due to strong electric fields. The calculated meridional thermospheric wind Vnx is more equatorward and on disturbed days. The strong Vnx (~90 m/s northward) on April 11 can explain the fast recovery of the local thermospheric parameters to the April 9 level. Calculations indicate that the ionosphere was molecular-ion-dominated up to 350 km on April 10, requiring a correction to the routinely derived Te(h), Ti(h), and Ne(h) radar profiles.¿ 1997 American Geophysical Union